Perplexity AI, a company rapidly expanding its AI search assistant onto Android smartphones through deals with major manufacturers, is facing serious questions about its application security. An audit published by mobile security firm Appknox concluded that Perplexity’s Android app, which reportedly boasts over 10 million downloads on Google Play, is riddled with severe vulnerabilities that expose users to significant risks, including data theft and account compromise.
The findings cast a shadow over the company’s simultaneous push into device ecosystems and its substantial venture capital fundraising efforts.
Fundamental Protections Appear Lacking
The Appknox report detailed a concerning list of security failures. Researchers found hardcoded secrets, such as API keys, embedded directly within the app’s code. This oversight could allow attackers who decompile the application to potentially extract these keys and gain unauthorized access to backend services.
The audit also identified improper Cross-Origin Resource Sharing (CORS) configurations using wildcard origins (`*`) in API responses. CORS is a browser security feature that restricts web pages from making requests to a different domain than the one that served the page; using a wildcard effectively disables this protection for Perplexity’s API, potentially allowing malicious websites to make unauthorized requests and extract user data.
Furthermore, the lack of SSL pinning – a technique where an application verifies the server’s identity by checking its certificate against a known copy – makes the app susceptible to Man-in-the-Middle (MitM) attacks, where communication between the app and its servers could be intercepted.
Appknox’s analysis went further, finding the application’s bytecode is not obfuscated, simplifying the process for attackers to reverse-engineer the app’s logic and discover additional weaknesses.
The report also noted the absence of checks for common debugging tools like Android Debug Bridge (ADB) or active developer options, potentially allowing malicious actors greater ease in manipulating the app’s behavior in controlled or compromised environments. Appknox stated these issues collectively make the app unsafe, advising users to uninstall it for the time being.
Deepseek Comparison Highlights Recurring Issues
The security firm drew direct parallels to a previous investigation into the Deepseek AI application, stating that Perplexity’s app exhibited all the flaws found there, plus five additional critical vulnerabilities unique to Perplexity.
Shared problems across both apps, according to Appknox, included categories such as unsecured network configurations, lack of SSL validation or certificate pinning, weak root detection, susceptibility to Tapjacking attacks (where users are tricked into clicking something unintended), exposure to the StrandHogg vulnerability (allowing UI hijacking), and the Janus vulnerability (allowing malicious code injection into legitimate app updates).
“Our testing highlights critical vulnerabilities in Perplexity AI that expose users to a variety of risks, including data theft, reverse engineering, and exploitation,” Appknox CEO and Co-Founder Subho Halder commented in the report. He added, “It’s crucial for the developers to address these issues swiftly. In the meantime, users should be cautious about using the app, particularly for sensitive activities.”
Expansion Meets Security Concerns
These revelations arrive at a moment of high momentum for Perplexity. The company recently finalized a deal with Motorola and is pursuing discussions with Samsung to integrate its AI assistant, launched for Android in January, directly onto Galaxy devices as an alternative to Google Gemini.
This strategy aims to place Perplexity’s multimodal assistant – which allows text, voice, and camera interaction – directly in front of millions of users via its main Android app.
This context makes the security findings particularly alarming, as flaws could affect users who haven’t explicitly chosen to install the app. The Samsung talks proceed despite Samsung’s deep existing partnership with Google on AI features, although Samsung’s venture arm, Samsung NEXT, is a prior investor in Perplexity and reportedly considering further investment.
A Pattern of Questions Amidst Growth
This push for device integration is backed by considerable financial activity, with Perplexity reportedly seeking $500 million to $1 billion in a funding round targeting an $18 billion valuation, supported by a claimed $100 million annual recurring revenue.
The company has also been actively broadening its offerings, launching its Sonar API for enterprise use, a “Deep Research” mode, and even an in-search e-commerce capability via a partnership with Firmly announced in late March. This handling of potentially sensitive payment and shipping information adds another layer of concern regarding the app’s security foundation.
The Appknox audit isn’t the first time Perplexity has faced scrutiny over its practices. Forbes and Wired have accused the company of aggressively scraping web content for its AI summaries, allegedly ignoring publisher instructions (robots.txt) and proper attribution, leading Amazon Web Services to reportedly investigate potential terms of service violations.
Perplexity denied these allegations at the time. Earlier, in January, CEO Aravind Srinivas addressed privacy concerns about using the Chinese-developed DeepSeek model, assuring users that Perplexity hosted it on US/EU servers and stating, “None of your data goes to China.”
While the company maintains a dedicated security page outlining policies like data segregation, access controls via AWS IAM, and use of tools like Cloudflare and Wiz, the Appknox findings suggest potential gaps between stated infrastructure security and the practical security implementation within the Android application itself.
